A repository with homework that I did for fast.ai course v3 (https://course.fast.ai/index.html)
I am currently learning about deep learning, hoping to apply it to biodiversity sciences pretty soon. For those who want to do the same, I recommend the amazing course Practical Deep Learning for Coders. Here you will find some of the homework that I did for the course:
Here I just adapted some code seen in class to classify dog and cat breeds to classify an insect order based on its egg image. Egg images come from a paper I recently wrote with Sam Church, Seth Donoughe and Cassandra Extavour. More specifically, from cropped images available in the egg visualization tool. This was really my first time coding a neural network and I did not expect much, since images include line drawings, photographs and SEMs. Surprisingly, the classifier did a reasonably good job at guessing the insect order based on an egg image.
Church SH, Donoughe S, de Medeiros BAS, et al. 2018. Insect egg size and shape evolve with ecology, not developmental rate. bioRxiv: 471946. doi:10.1101/471946.
Church SH, Donoughe S, de Medeiros BAS, et al. 2018. A database of egg size and shape from more than 6,700 insect species. bioRxiv: 471953. doi:10.1101/471953.